Regulatory Capture by Sophistication

Transcription

1 Regulatory Capture by Sophistication Hendrik Hakenes University of Bonn and CEPR Isabel Schnabel University of Bonn and CEPR ACPR Academic Conference Banque de France, Paris, December 2017

2 1. Introduction Why did financial regulation not prevent the financial crisis? Why did the increase in sophistication in financial regulation (e. g., the transition from Basel I to Basel II) not go along with safer financial systems but rather with a decrease in unweighted capital ratios to levels of 2 percent? Is a further increase in the sophistication of financial regulation desirable, or should less sophisticated regulatory measure play a bigger role (such as the leverage ratio, liquidity coverage ratio, net stable funding ratio,... )? Regulatory Capture by Sophistication Hendrik Hakenes & Isabel Schnabel 2/36

3 Regulatory Capture By Sophistication Our View Here: regulatory capture by sophistication micro-founded model with fully rational agents Conflict of interest: banks never want to be regulated, but sometimes regulation is efficient ( persuasion game ) Banks try to persuade the regulator to abstain from regulation by presenting sophisticated arguments If the regulators are not sufficiently sophisticated, they may not understand the arguments and rubber-stamp banks because they are not willing to admit that they do not understand the argument Reason: career concerns (shortcut for reputational concerns) Regulatory Capture by Sophistication Hendrik Hakenes & Isabel Schnabel 3/36

4 Examples 1. Bank Closure Bank in distress is to be bailed out or bailed in Optimal decision depends on the bank s state Banker knows best, situation can be complex 2. Supervision A bank is supervised (in Germany, by BaFin, Bundesbank and EBA; en France, par l autorité des marchés financiers,... ) Basel II and above Pillar 2 gives the supervisor discretion The supervisor can accept or reject a risk model (in IRB),... Complex situation, bank might know the quality of its models better, can better assess its situation 3. Regulatory Process Basel Committee on Banking Supervision decides upon increasing capital requirements Optimal decision depends on the banking system s state and structure Complex situation, bank organizations have valuable info 4. Non-bank examples Regulatory Capture by Sophistication Hendrik Hakenes & Isabel Schnabel 4/36

5 Literature on Regulatory Capture Stigler (BellJE, 1971) The Theory of Economic Regulation : seminal paper for the theory of regulatory capture, without ever using this phrase Laffont/Tirole (QJE, 1991) The Politics of Government Decision-Making: A Theory of Regulatory Capture : influential paper on regulatory capture with economic model Grossman/Helpman (2001) Special Interest Politics : excellent overview of the theory of lobbying But: no models on the role of sophistication, which is key in a complex environment such as banking regulation Regulatory Capture by Sophistication Hendrik Hakenes & Isabel Schnabel 5/36

6 Martin Hellwig (2010) Capital Regulation after the Crisis: Business as Usual? Chapter 3: Regulatory Capture by Sophistication: A Brief History of Capital Regulation When the model-based approach to capital regulation was introduced, however, the regulatory community was so impressed with the sophistication of recently developed techniques of risk assessment and risk management of banks that they lost sight of the fact that the sophistication of risk modeling does not eliminate the governance problem which results from the discrepancy between the private interests of the bank s managers and the public interest in financial stability. Regulatory Capture by Sophistication Hendrik Hakenes & Isabel Schnabel 6/36

7 Main Results In a pooling equilibrium, unsophisticated regulators pretend to understand sophisticated arguments and rubber-stamp bad banks Especially bad banks try to cheat the regulator When banks become more sophisticated it becomes easier to make complex arguments... regulators must raise the standards for arguments... an unsophisticated regulator understands fewer arguments... banks cheat more and get through with it Main result I: ignoring the bankers arguments would be a dominant strategy Main result II: the more sophisticated the banks are (relative to regulators), the worse is the regulatory outcome Regulatory Capture by Sophistication Hendrik Hakenes & Isabel Schnabel 7/36

8 Regulatory Capture by Sophistication 1. Introduction 2. The Model Environment Argumentation and Sophistication The Bank as an Urn Time Line 3. Equilibrium The Banker s Strategy The H-Regulator s Strategy Comparative Statics 4. Conclusion Regulatory Capture by Sophistication Hendrik Hakenes & Isabel Schnabel 8/36

9 2. The Model the Bank Bank with probability p of Success Probability 1 p of Distress Unknown type of the bank: unknown p Unobservable p is distributed with density f (p), support [0, 1] Alternative interpretation: several banks, some better some worse Expected type: E[p] = 1 0 p f (p) dp Stricter regulation causes cost C B for the bank Regulatory Capture by Sophistication Hendrik Hakenes & Isabel Schnabel 9/36

10 The Bank f p E p p Regulatory Capture by Sophistication Hendrik Hakenes & Isabel Schnabel 10/36

11 The Regulator The regulator weighs the cost of regulation against the expected cost from a financial crisis Social cost of regulation: C regulation Reduction in aggregate loan volume, costs CB for the bank Costs of distress for an unregulated bank: C distress > C regulation Regulated bank: no costs of distress (or even no distress) There is a critical bank type p that should just be regulated, with C regulation = (1 p) C distress Assumption: ex ante, regulation would be efficient, C regulation < (1 E[p]) C distress If the bank wants to remain unregulated, it must convince the regulator Regulatory Capture by Sophistication Hendrik Hakenes & Isabel Schnabel 11/36

12 The Bank f p E p p p E[p] < p: a priori, the regulator prefers to regulate The banker must put forward arguments against regulation Regulatory Capture by Sophistication Hendrik Hakenes & Isabel Schnabel 12/36

13 The Bank as an Urn Nature draws the final outcome (S or D), like from an urn Regulatory Capture by Sophistication Hendrik Hakenes & Isabel Schnabel 13/36

14 The Bank as an Urn S Nature draws an S-ball: success Regulatory Capture by Sophistication Hendrik Hakenes & Isabel Schnabel 13/36

15 The Bank as an Urn D Nature draws a D-ball: distress Regulatory Capture by Sophistication Hendrik Hakenes & Isabel Schnabel 13/36

16 The Bank as an Urn S S D S S S D S D S Relatively good bank: high p, many S-balls Regulatory Capture by Sophistication Hendrik Hakenes & Isabel Schnabel 13/36

17 The Bank as an Urn S D D D S S D D D D Relatively bad bank: low p, many D-balls Regulatory Capture by Sophistication Hendrik Hakenes & Isabel Schnabel 13/36

18 The Bank as an Urn S Argument: the banker presents the regulator an S-ball The banker produces arguments at cost c (test-draw from urn) Can decide whether to show the regulator, or draw again Regulatory Capture by Sophistication Hendrik Hakenes & Isabel Schnabel 13/36

19 The Bank as an Urn S 0.3 S 0.8 D 0.7 S 0.1 S 0.5 S D S D S 0.6 Complexity: potential arguments have numbers Regulatory Capture by Sophistication Hendrik Hakenes & Isabel Schnabel 13/36

20 The Bank as an Urn S 0.3? 0.8 S? S 0.5? D? D S 0.6 Sophistication: complex arguments are not understood (here: κ = 0.6) Regulatory Capture by Sophistication Hendrik Hakenes & Isabel Schnabel 13/36

21 The Bank as an Urn S 0.6 The banker presents the regulator an S-ball with low complexity If the argument were wrong (D-ball), the regulator could rebut it Regulatory Capture by Sophistication Hendrik Hakenes & Isabel Schnabel 14/36

22 The Bank as an Urn? 0.9 Possible: the banker presents an argument too complex for the regulator understand The regulator cannot rebut it, he can only admit that he does not understand Regulatory Capture by Sophistication Hendrik Hakenes & Isabel Schnabel 14/36

23 Argumentation Final state: Y {Success, Distress} The banker can spend c to produce a potential argument Two components, 1. a test draw Y i for the final state, either S (probability p) or D (probability 1 p), independent from the final state 2. the argument s complexity κ i, drawn from a uniform distribution The banker can produce as many arguments as he likes, all stochastically independent, {Y 1, κ 1 }, {Y 2, κ 2 }, {Y 3, κ 3 },... Producing arguments is unobservable to the regulator Regulatory Capture by Sophistication Hendrik Hakenes & Isabel Schnabel 15/36

24 Sophistication Agents differ by their degree of sophistication ( κ) Banker: κ B can make an argument i only if κb > κ i Otherwise: potential argument too complex useless Higher κb : smarter banker κ B is public information Regulator: two types Sophistication either κh (high type, fraction ϑ) or κ L (low type, fraction 1 ϑ) Argument with κ i > κ H (or κ L, depending on type): the regulator cannot observe Y i, does not understand the argument An unsophisticated regulator wants to appear smart Rationale: later wage negotiation, where productivity depends on perceived type (H or L) Regulatory Capture by Sophistication Hendrik Hakenes & Isabel Schnabel 16/36

25 Properties of the Persuasion Technology The smarter the banker, the cheaper it is for him to find an S-ball Interesting range: κ L < κ B < κ H, the banker can be smarter than the regulator, or vice versa Smarter banker: can make arguments that the regulator cannot understand Smarter regulator: understands everything the banker argues A single argument implies some learning (Bayesian updating), but no complete revelation of the bank s type Regulatory Capture by Sophistication Hendrik Hakenes & Isabel Schnabel 17/36

26 Time Line t = 0 Nature draws the bank s type (p) and the regulator s type (H or L) The regulator announces which arguments will convince him The banker decides whether to produce arguments and draws test realizations The banker shows one argument to the regulator (or not) Regulator decides whether to regulate the bank or not t = 1 Nature draws the outcome (Success, Distress) If regulation: costs C regulation If distress & unregulated bank: cost C distress t = 2 Epilogue: Regulator gets job offer, wage depends on perceived type Equilibrium concept: Perfect Bayesian with Intuitive Criterion Regulatory Capture by Sophistication Hendrik Hakenes & Isabel Schnabel 18/36

27 Regulatory Capture by Sophistication 1. Introduction 2. The Model Environment Argumentation and Sophistication The Bank as an Urn Time Line 3. Equilibrium The Banker s Strategy The H-Regulator s Strategy Comparative Statics 4. Conclusion Regulatory Capture by Sophistication Hendrik Hakenes & Isabel Schnabel 19/36

28 Equilibrium Properties Pooling: The unsophisticated L-regulator must follow the strategy of the smart H-regulator Even if he does not understand, he will accept arguments that the smart H-type would also accept The smart H-regulator accepts only a subset of arguments Argumentation must not be too cheap, otherwise Bayesian learning too weak to be convincing Regulator can increase costs by demanding higher sophistication The smart H-regulator prefers arguments that are too complex for the unsophisticated L-type (higher reputation gain possible) Result: minimum level κ 0 of argument complexity Valid arguments: S-balls with κ [κ 0, κ B ] Regulatory Capture by Sophistication Hendrik Hakenes & Isabel Schnabel 20/36

29 Possible Constellations of Sophistication acceptable arguments (with endogenous) possible arguments for a banker 0 1 understandable arguments for an L regulator argument s complexity understandable arguments for an H regulator Regulatory Capture by Sophistication Hendrik Hakenes & Isabel Schnabel 21/36

30 The Banker s Strategy If the banker draws an argument he cannot use (he does not understand it or it is not convincing), he will keep on drawing Three Possible Strategies Honest strategy: Produce arguments, present an argument only if it is valid (S and κ [κ 0, κ B ]) Advantage: be sure to convince the regulator Cheating strategy: Produce arguments, present an argument if it is either valid (S and κ [κ 0, κ B ]) or incorrect, but too complex for an unsophisticated regulator (D and κ ( κ L, κ B ]) Advantage: lower expected cost of argumentation Disadvantage: smart regulator will detect the cheating bank Zero strategy: Do nothing, accept to be regulated Advantage: no cost of argumentation Regulatory Capture by Sophistication Hendrik Hakenes & Isabel Schnabel 22/36

31 Honest Strategy, Zero Strategy Honest strategy: Draw until you find an S-argument in [κ 0, κ B ] Expected cost: c p ( κ B κ 0 ) Expected benefit: C B (avoided cost of regulation) Zero strategy: expected cost 0, expected benefit 0 Regulatory Capture by Sophistication Hendrik Hakenes & Isabel Schnabel 23/36

32 Cheating Strategy Draw until S-argument in [κ 0, κ B ] or D-argument in [κ 0, κ L ] Expected cost (if κ 0 κ L ): c p ( κ B κ 0 ) + (1 p) ( κ B κ L ) Probability of an S-argument, eventually: p ( κ B κ 0 ) p ( κ B κ 0 ) + (1 p) ( κ B κ L ) (1) S-argument no regulation D-argument regulation if regulator is smart (probability ϑ) Expected benefit: p ( κ B κ 0 ) + (1 p) ( κ B κ L )(1 ϑ) p ( κ B κ 0 ) + (1 p) ( κ B κ L ) C B Regulatory Capture by Sophistication Hendrik Hakenes & Isabel Schnabel 24/36

33 The Banker s Decision The banker prefers the honest strategy if p p honest with p honest = c C B ϑ ( κ B κ 0 ) The banker prefers cheating if p [p cheat, p honest ) with p cheat = c/c B (1 ϑ) ( κ B κ L ) ( κ B κ 0 ) (1 ϑ) ( κ B κ L ) The banker prefers to do nothing (zero) if p < p cheat Regulatory Capture by Sophistication Hendrik Hakenes & Isabel Schnabel 25/36

34 The Banker s Strategies for given set of convincing arguments f p p cheat p honest 0.5 do nothing chea ting honest p Regulatory Capture by Sophistication Hendrik Hakenes & Isabel Schnabel 26/36

35 Probability of Presenting an S-Argument Pr y p cheat p honest p For a cheating banker, the probability is as in (1), p ( κ B κ 0 ) p ( κ B κ 0 ) + (1 p) ( κ B κ L ) Regulatory Capture by Sophistication Hendrik Hakenes & Isabel Schnabel 27/36

36 The H-Regulator s Strategy Pr y E H p y 1 p Here: regulator has set κ (not in the pic) Consequence: p honest = and p cheat = (dashed) Bayes after an S-argument: E[p S] = (green bar) Regulatory Capture by Sophistication Hendrik Hakenes & Isabel Schnabel 28/36

37 The H-Regulator s Strategy Bayesian Rule phonest p ( κ B κ 0 ) p f (p) dp p cheat p ( κ B κ 0 ) + (1 p) ( κ B κ L ) 1 + p f (p) dp p E[p S] = honest phonest p ( κ B κ 0 ) f (p) dp p cheat p ( κ B κ 0 ) + (1 p) ( κ B κ L ) 1 + f (p) dp p honest = (this example) Regulatory Capture by Sophistication Hendrik Hakenes & Isabel Schnabel 29/36

38 The H-Regulator s Strategy Critical p: p = 1 C regulation C distress In equilibrium, minimum complexity κ 0 is adjusted such that p = E[p S] The H-regulator is only just convinced after an S-argument Cannot raise κ 0 any higher cannot ignore a convincing argument (Intuitive Criterion) For a lower κ 0, it would not be optimal to drop regulation after an S-argument Example: C regulation = 1 and C distress = 3.33 Then p = 0.7, and consequently κ 0 = Regulatory Capture by Sophistication Hendrik Hakenes & Isabel Schnabel 30/36

39 Equilibrium Quality of Decision Expected cost of the H-regulator Some banks do not present any argument, or are caught cheating: C regulation = 1 Other banks (high p) present a convincing argument, some because they are lucky in cheating For all banks with a good argument, the regulator is indifferent between regulating or not The expected cost of C distress is also = 1 Expected cost of the L-regulator Banks always get away with cheating Worse regulatory decision, expected cost > 1 Consequence: In the aggregate, listening to the banker deteriorates the regulatory outcome Rule-based regulation better than discretion! Regulatory Capture by Sophistication Hendrik Hakenes & Isabel Schnabel 31/36

40 No Easy Way Out Why can t the regulator accept the simplest arguments only? A smart regulator wants to show that he is smart Accepting sophisticated arguments increases the probability that unsophisticated regulators take wrong decisions, thus increasing the reputation of smart regulators who do not take wrong decisions (or, at least, with lower probability) Smart regulators set the agenda, unsophisticated ones must mimic Endogenously, argumentation becomes sophisticated If taking into account arguments reduces welfare, why does the regulator listen at all? The regulator cannot commit not to listen: he is open to information, and he must update his beliefs based on this information If the regulator has discretion, he will endogenously use it to the worse Regulatory Capture by Sophistication Hendrik Hakenes & Isabel Schnabel 32/36

41 Comparative Statics w. r. t. κ B Assume the bank gets smarter (increase in κ B ) Cheaper for banks to make arguments To keep his expected E[p] constant, the H-regulator must increase κ 0 For an L-regulator, the range of understandable arguments ( κ L κ 0 ) decreases Additional incentive for banks to choose cheating strategy Average quality of regulatory decision drops If κ B κ L + c C B (1 ϑ), then p cheat = 0: the argument becomes completely uninformative for the L-regulator Regulatory Capture by Sophistication Hendrik Hakenes & Isabel Schnabel 33/36

42 Comparative Statics w. r. t. ϑ Consider a higher fraction ϑ of smart regulators First positive effect: more banks are caught cheating Cheating strategy less beneficial for banks Smaller fraction of banks cheat Second positive effect: disutility of L-regulator from cheating banks is reduced double whammy Similar comparative static from increase in κ L No effect from increase in κ H Regulatory Capture by Sophistication Hendrik Hakenes & Isabel Schnabel 34/36

43 4. Conclusion Model has shown that banks can capture the regulator with their sophistication in equilibrium Three important ingredients: urn model of persuasion technology varying sophistication of banks and regulators career concerns Increase in sophistication in the banking sector was shown to be harmful (if it exceeds the minimum for regulators) Regulator can be captured more easily if banks are highly sophisticated Discretion reduces welfare An increase in the sophistication of financial regulation is not sufficient to provide for stable banking systems Regulatory Capture by Sophistication Hendrik Hakenes & Isabel Schnabel 35/36

44 Policy Implications Potential Solutions Relying on strict rules can improve regulation Less sophisticated regulation regime (e. g., standard approach under Basel Accord or non-risk-weighted capital ratios) may be preferable Leverage ratio, Liquidity Coverage Ratio, NSFR,... Increase sophistication of regulators relative to banks?? Immunize regulators against the consequences of perceived low sophistication Tenure? Groups of supervisors to increase anonymity? Regulatory Capture by Sophistication Hendrik Hakenes & Isabel Schnabel 36/36

45 Backup Reputation I True utility function of the regulator: U = λ ( χ C regulation + (1 χ) (1 p) C distress ) + w where w [L; H] depends on the expected type of the regulator, based on public information, and χ = 1 if regulation in place, otherwise χ = 0 Fraction ϑ of high types, hence ex ante w = ϑ H + (1 ϑ) L Assumption: regulated banks can still get into distress, but costs avoided Case 0: the banker does not present an argument No new information, χ = 1 U 0 = λ C regulation + ϑ H + (1 ϑ) L Regulatory Capture by Sophistication Hendrik Hakenes & Isabel Schnabel 37/36

46 Backup Reputation II Case 1: banker presents an argument in the range κ [κ 0, κ B ] Case 1.a: κ [κ 0, κ L ] Case 1.a.i: regulator accepts the argument = χ = 0 Banker must have chosen honest or cheating strategy Both H- and L-regulators do this w = ϑ H + (1 ϑ) L Case 1.a.ii: regulator rejects the argument = χ = 1 Out of equilibrium Case 1.b: κ [ κ L, κ B ] Case 1.b.i: regulator accepts the argument = χ = 0 Interesting case! Case 1.b.ii: regulator rejects the argument = χ = 1 Jackpot: only H-regulators can do this w = H Banker must have chosen cheating strategy, p [p cheat ; p honest ] Regulatory Capture by Sophistication Hendrik Hakenes & Isabel Schnabel 38/36

47 κ [ κ L, κ B ], Regulator Accepts I Public perspective: could be L-regulator who did not understand the argument, or H-regulator after S No distress case: w = phonest p ( κ B κ L ) + 0 (1 p) ( κ B κ L ) p f (p) dp p cheat p ( κ B κ 0 ) + (1 p) ( κ B κ L ) 1 κ B κ L + p f (p) dp p honest κ B κ 0 phonest p ( κ B κ L ) + (1 p) ( κ B κ L ) p f (p) dp p cheat p ( κ B κ 0 ) + (1 p) ( κ B κ L ) 1 + κ B κ L p f (p) dp p honest κ B κ 0 Regulatory Capture by Sophistication Hendrik Hakenes & Isabel Schnabel 39/36

48 κ [ κ L, κ B ], Regulator Accepts II Distress case: w = phonest p ( κ B κ L ) + 0 (1 p) ( κ B κ L ) (1 p) f (p) dp p cheat p ( κ B κ 0 ) + (1 p) ( κ B κ L ) 1 κ B κ L + (1 p) f (p) dp p honest κ B κ 0 phonest p ( κ B κ L ) + (1 p) ( κ B κ L ) (1 p) f (p) dp p cheat p ( κ B κ 0 ) + (1 p) ( κ B κ L ) 1 + κ B κ L (1 p) f (p) dp p honest κ B κ 0 Assumption slide 16 (unsophisticated regulator wants to appear smart) is consistent if λ such that expected disutility (suboptimal regulation) < expected utility (higher wage) Yet to be shown: especially if E[1 p] is small (banking) Regulatory Capture by Sophistication Hendrik Hakenes & Isabel Schnabel 40/36